This project was developed for the Machine Learning (SSC0276) course at the University of São Paulo (USP) by:

• Alice Valença De Lorenci - 11200289
• Gabriel Soares Gama - 10716511
• Marcos Antonio Victor Arce - 10684621

The MRI dataset used was kindly provided by ADNI and anyone wishing to reproduce our results may apply for access to the data.

The code is organized in the folders data, models, mobileNet and preprocessing, and the relevant files are:

• preprocessing:
  • preprocessing.ipynb: preprocessing flow applied to the data
  • feature_extraction.ipynb: feature extraction using a CNN pre-trained on the ImageNet database
  • load_dataframe.ipynb: instructions on how to load a dataframe with the extracted features
• data:
  • features.npz: features extracted from the MRI images
  • name_class.csv: classes of the MRI images
• models:
  • KNN.ipynb: KNN model
  • MultilayerPerceptron.ipynb: multilayer perceptron model
  • SVM.ipynb: SVM model
• mobileNet:
  • dataloader.py: loads the images in the required format for MobileNetV2
  • train.py: MobileNetV2 model
  • eval.py: evaluation metrics